{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/sodeep-a-sorting-deep-net-to-learn-ranking","title":"SoDeep: a Sorting Deep net to learn ranking loss surrogates","arxiv_id":"1904.04272","date":"2019-04-08","proceeding":"CVPR 2019 6","authors":["Martin Engilberge","Louis Chevallier","Patrick Pérez","Matthieu Cord"],"abstract":"Several tasks in machine learning are evaluated using non-differentiable\nmetrics such as mean average precision or Spearman correlation. However, their\nnon-differentiability prevents from using them as objective functions in a\nlearning framework. Surrogate and relaxation methods exist but tend to be\nspecific to a given metric.\n  In the present work, we introduce a new method to learn approximations of\nsuch non-differentiable objective functions. Our approach is based on a deep\narchitecture that approximates the sorting of arbitrary sets of scores. It is\ntrained virtually for free using synthetic data. This sorting deep (SoDeep) net\ncan then be combined in a plug-and-play manner with existing deep\narchitectures. We demonstrate the interest of our approach in three different\ntasks that require ranking: Cross-modal text-image retrieval, multi-label image\nclassification and visual memorability ranking. Our approach yields very\ncompetitive results on these three tasks, which validates the merit and the\nflexibility of SoDeep as a proxy for sorting operation in ranking-based losses.","url_abs":"http://arxiv.org/abs/1904.04272v1","url_pdf":"http://arxiv.org/pdf/1904.04272v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"sodeep-a-sorting-deep-net-to-learn-ranking","repo_url":"https://github.com/technicolor-research/sodeep","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause-Clear"}}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"multi-label-image-classification","task_name":"Multi-Label Image Classification"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1904.04272","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}